OmniMAE: Single Model Masked Pretraining on Images and Videos | IEEE Conference Publication | IEEE Xplore